1

The Application of Carsons Equation to the

Steady-State Analysis of Distribution Feeders
W. H. Kersting, Life Fellow, IEEE, and R. K. Green, Life Senior Member, IEEE

AbstractOne of the primary purposes of performing

the steady-state analysis of a distribution feeder is to
determine the voltages at every node. Because these
voltages are a function of the line voltage drops it is critical
that the line impedances used are as exact as possible. In
1926 John Carson developed equations that would
determine the self and mutual impedances of any number
of overhead or underground conductors taking into
account the effect of ground [1]. In recent years the
application of Carsons equation has become the standard
for the computation of line impedances. Because Carsons
equation results in an infinite series, approximations have
been made to ease in the computation of the impedances.
The purpose of this paper is to investigate some of the
more common approximations and determine what, if any,
errors are made.
Index Termsline impedance calculations, distribution
system analysis, overhead line impedances, underground
impedances, Kron reduction, resistivity
I. INTRODUCTION
When Carsons paper first appeared in 1926 it was not met
with a lot of enthusiasm because of the complexity of the
equations and the only tool for computations was the slide
rule. With the advent of the digital computer the equations
could now be incorporated into the computation of the
impedances of overhead lines [2]. The original equations
resulted in an infinite integral [3]. Because of this various
approximations have been made in order to make the equations
more useful.
In 1991 a paper giving the data for four distribution system
test feeders was published [4]. The purpose of the test feeders
was to give software developers a common set of data that
could be used to verify the correctness of their programs.
Since then the original four test feeders along with additional
special purpose test feeders have been made available on the
IEEE website [5].
As a result of developers and students using the test feeders
to develop programs the question has come up as to what
method should be used to compute the overhead and
underground line impedances. Casons equations are used by

many but there have been several approximations used to

simplify the calculations. For example, in Reference [7], the
modified Carsons equations are developed. While these
equations are simple and straight forward to apply there is
some concern among developers if the full version of Carsons
equations should be used. What are the errors introduced
using the modified equations? This paper will address that
question by applying the full and modified equations to a
three-phase overhead line, a three-phase concentric neutral
underground line and a four wire quadraplex cable. Since the
concern is for the computation of line impedances for the
steady-state analysis a frequency of 60 Hz will be used. The
value of the earth resistivity will also be addressed.
II. CARSONS EQUATIONS
Carsons equations for the self-impedance with earth return
and mutual impedances with common earth return are [1]:

Sii
zii ri 4 P G j 2 G xi ln
2Q
Radiusi

Sij

zij 4 P G j 2 G ln
2 Q
Dij

where: i j 1, 2 ncond
ncond number of conductors
G 0.1609347 10
2 f

Equations 1 give the impedances in /mile.

The spacing between conductors and their images used in
Equations 1 are shown in Figure 1.
i
Dij

ij
Sii

Sij

j'

W. H. Kersting is a consultant to Milsoft Utility Solutions, Abilene, TX,

(e-mail: bjkersting@zianet.com).
R.K. Green is with Garland Power & Light, Garland, TX, (e-mail:
bobg@gplops.org).

(1)

i'

Figure 1 - Spacings

In Equations 1:

Pij

Radiusi
xi ln

GMRi
where: Radiusi conductor radius in ft.

(2)

1
zii ri 4 P G j 2 G ln
ln Sii 2 Q
GMR
i

1
zij 4 P G j 2 G ln
ln Sij 2 Q
Dij

(3)
The difficulty in applying Equations 3 lies in the terms of P
and Q. For purposes of this paper, P and Q are defined by
Equations 4 and 5 [3],[6]. In the rest of this paper, these will
be referred to as the full Carsons equations.

kij 2
16

1
3 2

kij cos ij
k ij 3

ij sin ij

kij 2 cos 2 ij

2 45

0.6728 ln 2

16

cos 3 ij

kij 4
1536

kij

cos 4 ij

(4)
kij
1
2
1
ln

kij ij
cos 2 ij
2
kij 3 2
64
2

Qij 0.0386

kij 3
45 2

cos 3 ij

kij 4
384

ij sin 4 ij

2
1
ln
2 kij

1
zii 0.09530 ri j 0.12134 ln
7.93402
GMR
i

1
zij 0.09530 j 0.12134 ln
7.93402
Dij

(5)

In applying both the full and the modified Carsons

equations an ncond x ncond primitive impedance matrix is
developed. The terms in this matrix include the selfimpedance of each conductor and the mutual impedances
between all conductors. For a four wire line the matrix will be
4 x 4. Included in each term, thanks to Carson, will be the
effect of the self-impedance of ground and the mutual
impedances between each conductor and ground. Since the
neutral is grounded (the LG voltage is zero), the 4 x 4
primitive impedance matrix can be reduced to a 3 x 3 phase
impedance matrix. The 4 x 4 primitive impedance matrix is
partitioned between the 3rd and 4th rows and columns to form:

zin
znn

(9)

Applying the Kron reduction the 3 x 3 phase impedance

matrix is determined by:

zabc zij zin znn1 zjn

In Equations 4 and 5:

Where:

(8)

III. KRON REDUCTION

zij
z primitive

znj

kij 4 cos 4 ij 2

ln
1.0895

384
kij

kij 8.565 10

(7)

The modified Carsons equations assuming a frequency of 60

Hz and a resistivity of 100 -meters are:

GMRi = conductor geometric mean radiusl in ft.

Qij 0.0386

Substitute Equation 2 into Equations 1 and simplify:

Pij

(10)

IV. FOUR WIRE OH LINE

4

Sij

f frequency = 60 Hz
resistivity of earth in Ohm-meter

(6)

It is easy to see why there was reluctance in applying these

equations using a slide rule in 1926.
In Reference [7], the modified Carsons equations are
developed. This modification limits the number of terms used
in P and Q in Equations 3. For this modification the terms
used for P and Q are:

A four-wire overhead line is shown in Figure 2. The phase

impedance matrix will be developed for this line using the full
and modified Carsons equations. The purpose is to determine
how much of an error is made using the modified equations as
opposed to the full equations.
2.5'
b

4.5'
a

c
3.0'
4.0'

n
25.0'

Figure 2 Four Wire Overhead Line

Figure 2 represents configuration 400 in the IEEE Test

Feeders [5]. The phase and neutral conductors are 1/0 ACSR
6/1 with the following:

rdirt 0.09530

Conductor resistance = 1.12 /mile

GMR = 0.0045 ft.
Diameter = 0.398 inches

j1.6184
j 0.8515
j 0.7802
j 0.7865

0.0953
1.2153
0.0953
0.0953

j 0.8515
j1.6184
j 0.7266
j 0.7543

0.0953
0.0953
1.2153
0.0953

(15)

constant = 7.93402

The mean values of the dirt resistance and constant for the
full equations are:

Note in Figure 2 that the phase sequence is b-a-c and that all
impedance matrices that follow have units of /mile.
Applying the modified equations the primitive impedance
matrix in is:
1.2153

0.0953
zpmod
0.0953

0.0953

impedance matrix as seen in Equations 10 and 11. With that

in mind the mean value of the dirt resistance and the mean
value of the constant are used in what will be referred to as the
new modified equations. In the original modified equations
the dirt resistance and constant are:

j 0.7802
j 0.7266
j1.6184
j 0.7524

0.0953
0.0953
0.0953
1.2153

j 0.7865

j 0.7524
j 0.7674

j1.6184

rdirt 0.09327

(16)

constant = 7.95153

Using these mean values the new modified equations are;

1
zii 0.09327 ri j 0.12134 ln
7.95153
GMR
i

zij 0.09327 j 0.12134 ln

7.95153
Dij

(17)

(10)
Applying the full equations the primitive impedance matrix

When these equations are used the errors compared to the

full equations are:

is:
1.2132

0.0932
zp full
0.0932

0.0933

j1.6206
j 0.8537
j 0.7824
j 0.7886

0.0932
1.2132
0.0932
0.0933

j 0.8537
j1.6206
j 0.7288
j 0.7545

0.0932
0.0932
1.2132
0.0953

j 0.7824
j0.7288
j1.6206
j 0.7674

0.0933
0.0933
0.0953
1.2132

j 0.7886

j 0.7545
j 0.7674

j1.6206

(11)
In Equations 10 and 11 notice the small differences in the
real and imaginary components of the matrix elements.
The Kron reduction is applied to the modified and full
primitive impedance matrices to develop the phase impedance
matrices. Those matrices are:
1.3369 j1.3331 0.2102 j 0.5778 0.2132 j 0.5014

zmod 0.2102 j 0.5778 1.3239 j1.3557 0.2067 j 0.4591

0.2132 j 0.5014 0.2067 j 0.4591 1.3295 j1.3459
1.3366 j1.3346 0.2099 j 0.5793 0.2128 j 0.5029

z full 0.2099 j 0.5793 1.3235 j1.3573 0.2063 j 0.4606

0.2128 j 0.5029 0.2063 j 0.4606 1.3292 j1.3474

(12)

Imerror 0.2328 %

(18)

Imerror 0.0183 %

The errors have been reduced but it must be understood that

the mean values of the new dirt resistance and constant are a
function of the spacings between the conductors on the poles.
Change the spacings and the new dirt resistance and constant
will also change. Is it really worth the effort?

V. UNDERGROUND CONCENTRIC NEUTRAL CABLES

Configuration 425 in the IEEE Test Feeders is a three-phase
underground concentric neutral line with an additional neutral
conductor is shown in Figure 3.

(13)

In order to compare the two methods the mean error on both

the real and imaginary parts are computed to be:

Realerror 0.1334 %

Realerror 0.0342 %

6"

6"

6"

Figure 3 UG Concentric Neutral Line

(14)

This very small error is the result of the difference in

specified resistance of dirt and the constant in the imaginary
parts in the modified and full equations. Using the full
equations a slightly different value of the dirt resistance and
imaginary constant are computed for each term of the primitive

This line consists of 3 phase conductors, 3 equivalent

concentric neutral conductors and the additional neutral
conductor. In effect [7] there are 7 conductors which leads to
a 7x7 primitive impedance matrix.
As was done with the three-phase OH line, the modified and
full Carsons equations were used to compute the primitive

impedance matrices. The matrices are partitioned between the

3rd and 4th rows and columns. The Kron reduction is used to
develop the phase impedance matrices shown below.
1.1339 j 0.6453 0.3189 j 0.1770 0.2785 j 0.0859

zmod 0.3189 j 0.1770 1.1253 j 0.5895 0.2991 j 0.1327

0.2785 j 0.0859 0.2991 j 0.1327 1.0995 j 0.5518

(19)

1.1340 j 0.6454 0.3189 j 0.1770 0.2786 j 0.0850

z full 0.3189 j 0.1770 1.1254 j 0.5895 0.2991 j 0.1327

0.2786 j 0.0850 0.2991 j 0.1327 1.0995 j 0.5518

(20)

(21)

Imerror 0.0094 %

This small of an error does not warrant computing new dirt

resistance and constant values.
VI. QUADRAPLEX SECONDARY
Configuration 460 in the IEEE Test Feeders is a 4 wire
quadraplex cable used in many three-phase secondary circuits.
This cable is shown in Figure 4.

Imerror 0.0980 %

(24)

A correction for the dirt resistance and constant used in the

modified Carsons equations could be made as was done for
the OH line. However, with such a small error it doesnt seem
to be necessary.
VII. VOLTAGE DROP ERRORS
It has been demonstrated that there is very little error in
the phase impedance matrices for overhead, underground
and secondary lines. The question at this point is what
errors are made in the calculation of voltages and voltage
drops using the full and modified phase impedance
matrices? The overhead line will be used to demonstrate
the possible errors.
The IEEE 34 Node Test Feeder nominal LL voltage is
24.9 kV. However, the source is set at 1.05 per-unit which
means the source line-to-ground voltages are 15.095 kV.
From Node 800 to Node 808 there is an overhead line of
configuration 400. This configuration is the same as at the
start of this paper. The total distance from Node 800 to
Node 808 is 36,540 ft. There is a small distributed load
from Node 802 to Node 806 that will be ignored. The
complex power delivered to Node 808 is:
699 j143

SL 690 j103 kW+jkvar

591 j16

(25)

Using the modified Carsons equation phase impedance

matrix, the voltages at Node 808 are computed to be:

n
Figure 4 Four wire quadraplex cable
The phase conductors are 2/0 AA phase and the neutral
conductor is 2/0 ACSR. The thickness of the phase insulation
is 60 mils.
For this case the primitive impedance matrices are 4x4 and
the Kron reduction is used to compute the 3x3 phase
impedance matrices for the modified and full Carsons
equations.
1.2543 j 0.6031 0.4619 j 0.4936 0.4853 j 0.4490

zmod 0.4619 j 0.4936 1.2089 j 0.6901 0.4619 j 0.4936

0.4853 j 0.4490 0.4619 j 0.4936 1.2543 j 0.6031

(23)

There is very little error between these two matrices. To be

consistent the errors are:

Realerror 0.0688 %

As can be seen, there is very little difference between the

two phase impedance matrices. Just to be consistent the
percent errors are:

Realerror 0.0085 %

1.2547 j 0.6036 0.4623 j.4941 0.4857 j 0.4495

z full 0.4623 j.4941 1.2093 j 0.6906 0.4623 j 0.4941

0.4857 j 0.4495 0.4623 j 0.4941 1.2547 j 0.6036

(22)

14606.60/ .62

VLN mod808 14726.69/ 121.0

14801.37/119.2

(26)

The source voltages are:

15094.8/ 0

VLN800 15094.8/ 120

15094.8/120

(27)

For this operating condition the phase voltage drops from

Node 800 to Node 808 are?

vdropmod

VLN800 VLN808
VLN800

3.2344

100 2.3288 %
1.9441

(28)

For the same operating conditions the full Carsons equation

phase impedance matrix is used and the voltages at Node 808
are computed to be:
14606.46/ .62

VLN full808 14726.78/ 121.0

14801.42/119.2

(29)

VLN800 VLN808
VLN800

3.2353

100 2.3282 %
1.9437

vdrop full vdropmod

vdrop full

0.0278

100 0.0237 %
0.0166

VLN full808

0.0009

100 0.0006 %
0.0002

(34)

(35)

The load end voltages for 100 -meters were found to be:
(31)
14606.46/ .62

VLN full100 14726.78/ 121.0

14801.42/119.2

(36)

For this case the errors per phase between the 1000 and 100
-meter lines are:

The errors in the computed load end voltages are:

VLN full808 VLN mod808

1.3366 j1.3346 0.2099 j 0.5793 0.2128 j 0.5029

z100 full 0.2099 j 0.5793 1.3235 j1.3573 0.2063 j 0.4606

0.2128 j 0.5029 0.2063 j 0.4606 1.3292 j1.3474

14602.29/ .61

VLN full1000 14725.79/ 121.0

14806.68/119.2

Needless to say, the errors in using the modified vs. the full
phase matrices are insignificant.

V 808error

For ease of comparison, the phase impedance matrix for 100

-meter is:

(30)

Computing the error is voltage drops:

vdroperror

(33)

The phase impedance matrix of Equation 34 is used to

compute the load end voltages for the same line and load
conditions as in the previous section. The load end voltages
with the new phase impedance matrix are:

The phase voltage drops are computed to be:

vdrop full

1.3731 j1.3849 0.2459 j 0.6310 0.2491 j 0.5540

z1000 full 0.2459 j 0.6310 1.3591 j1.4105 0.2421 j 0.5131

0.2491 j 0.5540 0.2421 j 0.5131 1.3652 j1.3993

(32)

Equations 31 and 32 demonstrate that there are virtually no

errors between using the full or modified Carsons equations to
compute the phase impedance matrix. This is important since
there has been some concern between developers and students
using the test feeders what phase impedance they should be
using. Note that this paper has been addressing the phase
impedance matrix as applied to the steady-state power flow
studies.
VIII. THE EFFECT OF RESISTIVITY
For the computation of the phase impedance matrices the
resistivity of dirt has traditionally been assumed to be 100
-meter. The question is how sensitive are the impedances
and node voltages changed using the assumed resistivity?
Using the full Carsons equations with values of resistivity of
1000 -meter and 10 -meter (as compared to 100 -meter),
the effects on the phase impedances and load end voltages will
be investigated.
For a resistivity of 1000 -meter, the phase impedance
matrix is:

Verror

VLN full1000 VLN full100

VLN full1000

0.0286

100 0.0067 %
0.0355

(37)

When the resistivity is changed to 10 -meter the results are:

14611.24/ .64

VLN full10 14727.50/ 120.98

14795.80/119.22

Verror

VLN full10 VLN full100

VLN full10

(38)

0.0327

100 0.0049 %
0.0380

(39)

For both values of resistivity it is seen that the percentage

errors are very small when compared to the 100 -meter
resistivity.
It is possible to change the constant in the imaginary part of
the modified Carsons equations to take into account the
change in resistivity. The constant term is given by [7]:

6
constant =7.6786

1

ln
2
60

constant1000 7.6786
constant10 7.6786

1
1000
ln
9.08531
2
60

(40)

It should be repeated that this paper has been limited to the

application of the phase impedance matrices for the analysis of
steady-state conditions at 60 Hz.

1
10
ln 6.78272
2
60

X. REFERENCES
Applying these new constants in the modified Carsons
equations will yield the same very small percentage errors as
with the full equations.

IX. CONCLUSIONS
The primary purposes of this paper were to investigate how
the elements in the phase impedance matrix are affected by:
The number of terms used in Carsons equations
for:
o Overhead lines
o Concentric neutral underground lines
o Secondary quadraplex cable
The assumed value of resistivity
The phase impedance matrix was developed for the
following IEEE Test Feeder configurations:
Overhead - #400
Concentric neutral underground - #425
Secondary quadraplex - #460
For each configuration the phase impedance matrices were
computed using the modified Carsons equations from
reference [7] and the full Carsons equations from references
[3] and [6] assuming a resistivity of 100 -meters. A very
small error (less than 0.3%) was found in comparing the
individual elements in the two matrices. To test the error
introduced on the node voltages using the modified and full
equations a routine was developed to compute the node
voltages at the end of a three-phase overhead line serving an
unbalanced three-phase load. For this study it was found that
the error in the node voltages was of the order 0.0005 %.
The effect of resistivity was studied for the overhead line for
values of resistivity of 1000 and 10 -meters. These values
were used in computing the phase impedance matrices using
the full equations. The new matrices were used to compute the
load voltages for the unbalanced load. These voltages were
then compared to the voltages computed using the modified
equation that assumes 100 -meters. The errors were found to
be in the order of 0.03%. This very small error could be
minimized by changing the constant in the imaginary term of
the modified equations.
The IEEE Test Feeders are beginning to be used in many
papers submitted for presentation at conferences and/or
published in the transactions. There has been some concern
that the values of the phase impedance matrices used by the
authors might be the cause of paper results that do not match
the published results. It is hoped that this paper has
demonstrated that either the modified or full Carsons
equations may be used in the computation of the phase
impedance matrices.

1. John T. Carson, Wave Propagation in OverheadWires

with Ground Return, Bell System Technical Journal,
New York, NY, vol. t, 1926, pp. 539-54.
2. M.H. Hesse, Electromagnetic and Electrostatic
Transmission Parameters by Digital Computer,
IEEE Transactions (Power Apparatus and Systems),
vol. 82, pp 282-291, June 1963.
3. R.K. Green, J.L. Bowen, M.W. Chen, Calculation of
Transmission Line Constants, Southwest IEEE
Conference, conference paper, 1968.
4. IEEE Distribution Planning Working Group Report,
Radial distribution test feeders, IEEE Transactions
on Power Systems,. August 1991, Volume 6, Number
3, pp 975-985.
5. http://ewh.ieee.org/soc/pes/dsacom/testfeeders/index.html

6. Edith Clark, Circuit Analysis of A-C Power Systems,

John Wiley & Sons, New York, 1943.
7. W. H. Kersting, Distribution System Modeling and
Analysis, CRC Press, Boca Raton, FL, 2007.